Computers & Graphics 30 (2006) 423–431 Virtual Reality Interaction and Physical Simulation Approximate collision response using closest feature maps Thanh Giang Ã,1 , Carol O’Sullivan Image Synthesis Group, Department of Computer Science, Trinity College, Dublin, Ireland Abstract In this paper we consider the novel idea of closest feature maps (CFMs) applied to refinable collision response in order to address the potential issues and problems associated with over approximation of contact information for time-critical collision detection schemes that utilise sphere-tree bounding volume hierarchies. Existing solutions to time-critical collision handling can at times suffer from over- approximation of required contacting data, which may lead to undesirable or implausible physical response. Our CFM solution essentially utilises information for contact data approximation based on the underlying geometry of the colliding objects rather than potentially problematic properties of the contacting bounding volumes. The merits of the scheme lie in its simplicity and effectiveness to handle refinable collision data in an efficient manner and could quite easily be extended to incorporate other types of bounding volume hierarchies for interruptible collision handling. r 2006 Elsevier Ltd. All rights reserved. Keywords: Physically based modelling; Three dimensional graphics and realism; Time-critical dynamics; Collision detection and response 1. Introduction Obtaining useful information for collision resolution after a potential collision has been detected is vital for good plausible dynamic response. The traditional approach in a collision detection scheme that utilises bounding volume hierarchies (BVHs) such as sphere-trees, is to have each leaf of the tree contain the information about which polygon(s) it bounds. Upon traversal down to the leaf level signifying a positive collision, more expensive and exact schemes such as polygon-polygon, face-edge or similar intersection tests are further conducted. From these tests, useful collision information such as point(s) and normal(s) of contact are obtained for collision resolution. In such collision detection approaches, the BVHs are utilised as a detection accelera- tion technique only. For time-critical collision detection, such an approach may not be possible due to the uncertainty that we may ever reach the leaf level of our BVH. For such schemes, a BVH must be treated as inherently part of the entire collision handling process, whereby a quick definitive answer must be obtainable at any level of the hierarchy. The problem of what to do during interruption in a time- critical collision detection schema still very much remains an open question. To our knowledge, very few researchers to date have considered the question of the next step beyond interruption of collision detection. In this paper we build upon the work of Dingliana and O’Sullivan [1]. While their proposed contact resolution scheme works quite well, it can potentially provide over-approximations. This is most noticeable during cases of continuous contact and when we are forced to resolve collisions at higher levels of the bounding volume hierarchy, resulting in a detraction from the overall plausibility of the physical simulation. This paper proposes an improved refinable collision handling scheme that rectifies the potential drawbacks of Dingliana and O’Sullivan’s approximation scheme. To improve contact data approximation, we propose the use of closest feature maps (CFMs) to quickly approximate essential contact information at the polygonal level rather than using spherical geometry. The merits of the proposed scheme lie in its simplicity and robustness. Although we present the scheme here for spheres in a sphere-tree BVH, ARTICLE IN PRESS www.elsevier.com/locate/cag 0097-8493/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.cag.2006.02.019 Ã Corresponding author. E-mail addresses: thanhg@gmail.com (T. Giang), Carol.OSullivan@cs.tcd.ie (C. O’Sullivan). 1 This paper represents work done by the author while with the Image Synthesis Group, Trinity College Dublin as a Ph.D. Research Fellow.